Talk:Extended West Papuan languages

Is this article really necessary? Don't forget Occam's razor
kwami, I really appreciate you have taken the time to do such a good job on this page, but I had read M.Ross' paper some time ago and remembered well that it was far less conclusive than this article (and indeed this article's very existence) implies. Now, on looking it up once more, I can't help but feeling all the more sure about that. You see, if we were to create a full-fledged wiki article for every tentative (and nothing but tentative) grouping Ross included on that paper, there would be a bewildering number of barely-established quasi-families of languages which could mean even more confusion in the already sensitive and tricky topic of language classification (and in particular Papuan LC). To me, it seems quite clear that it was not Ross' intention at all to have a definitive and potentially 'encyclopedic' say on such a classification (besides, I'm not aware of any major work in the wake of this paper either confirming or discrediting it, which renders its theses somewhat loose and too poorly reviewed - at least yet - to be here.) Ross himself acknowledges (unlike, e.g., Greenberg, and maybe even Wurm) that the extremely limited amount of data he is dealing with is not really enough to 'officially' propose a language family, and admits, very reasonably, that he can't even claim to possess "probative evidence" for such genealogical groupings. So the only major advantadge over Greenberg's work here is that we have an author who knows he hasn't followed strictly the scientific method. He's adopting an absolutely cautious and almost agnostic attitude - and so should we.

Just illustrating how tentative (or "preliminary") and purely provisional all those classifications were, I'll quote here the most significant excerpts, in Ross' own words.

"The main purpose of this chapter is to provide a provisional listing of families of Papuan languages based largely on a single criterion, namely the pronoun sets of these languages.

(...) As well as ‘family’, the terms ‘subgroup’, ‘group’ and ‘microgroup’ are also used in this chapter. A ‘subgroup’, like a family, is a genealogically related set of languages, but one that forms part of a family. Its members are considered on present evidence to be more closely related ro each other than to other members of the family.

(...) Two strong caveats apply to Table 5. First, it is mostly based on an examination of pronoun forms and is only a statement of preliminary hypotheses about families. Where languages have resemblant pronoun sets, a family is proposed, but at least Steps 2, 3 and 4 of the comparative method need to be applied before we can say that we have probative evidence for it.

(...) There are far more statements above of the form "a historical linguist would consider it probable that a particular set of resemblant pronoun forms is due to genealogical relatedness" or "a historical linguist would consider it far more probable that a particular set of resemblant pronoun forms is due to genealogical relatedness than that it reflects borrowing". Here the statement is about the historical linguist´s strength of belief in a proposition. This is a Bayesian use of probability. The proposition itself is either true or false: either the set of resemblant pronoun forms is due to genealogical relatedness or it isn´t. Bayesian probability is a measure of strength of belief in an assertion.

(...) frequentist statistics has nothing to say about a statement like ‘a historical linguist would consider it probable that a particular set of resemblant pronoun forms is due to genealogical relatedness’, because the statement implies (and may posit) alternate hypotheses, e.g. that the set of resemblant pronoun forms is due to borrowing or to chance.

(...) The point of this note is rather that much of what has been expressed here in terms of probability and likelihood is an expression of strength of belief in certain propositions. Bayesian statistics has been coming into its own in recent years in association with forensic science, where the scientific expert is expected to offer the court an estimate of his strength of belief in a proposition offered in evidence. The science itself is (one hopes) empirically founded, but its application to a given hypothesis of guilt or innocence is not absolute. And so it is with the use of as evidence of language families." E.Cogoy 08:28, 16 April 2006 (UTC)


 * Fair enough, and the article could probably use some more caveats. However, I have created articles for all of Ross' family proposals (or at least all those he felt confident enough about to create a separate section heading for), and it's not overwhelming. We also have articles for proposals such as East Papuan and East New Guinea Highlands which have been debunked.


 * You may be right, but I'd like to put the info somewhere, and I thought a single article was preferable to repeating it with each constituent family. I suppose we could just say 'for more info, see West Papuan' in each of the others. kwami 12:45, 16 April 2006 (UTC)

Paleo-Sundic
Kwami I see you've included Usher's 'Paleo-Sundic' theory in this article which, as it's unpublished, counts as original research. Dougg 05:08, 17 April 2006 (UTC)
 * Okay, I've removed the term. The idea of a relationship, however, is hardly unique to Usher, so I'm leaving that in. kwami 05:30, 17 April 2006 (UTC)

The specific details used here to support that idea are exclusively his, and as far as I know they have not been published in any peer-reviewed source, so I'm removing them as OR. Also, Greenberg's pseudoscientific speculations have little or nothing to do with Ross' Extended West Papuan (at least directly), so I don't know what they are doing here. Ko'oy 23:49, 29 October 2007 (UTC)